If you could photograph a normal map, by painting something white then shining the colors on it you MIGHT be able to convert the deformation to mesh. Normally its the other way around, where you export normals from HD mesh.
Find a good way to go for normal maps to mesh and it will work to some degree. You won't get serious depth, but if your objects are kind flat it might work.
You'll also have to do serious edits on the normal map and the 3d model especially if you want to take these shots and convert them into a full 3d model, viewable from all angles.
edit:
If you want to try this with objects you don't want to paint white:
You might be better off shooting in black and white, using one white light.
Example: For a single normal map, with 3 camera shots from the same perspective you could have 3 different lighting directions by moving the light around each time. Then you use this data make a normal map, with each of the 3 shots being converted to the 3 specific colors that normal maps use. I could see that working for a car engine.
This sounds like fun.
edit: edit:
That's what this guy is doing!
http://zarria.net/nrmphoto/nrmphoto.html except you want to take it one step further and convert to a mesh object.
